which one is more matching for these two layouts?

[chang830]

Hi,
Would you pls. take a look at the attached two layouts? It is a 1:2 current mirrors.I hope to lay it with common-centeroid to make good matching.I used the two layout, but I don't know which one is more maching?

For the 1st one, it is matched for the Drain and source of the M1&M2.The signal line is also macthing.

But for the 2nd one,it is more like a common centroid maching.

Would anyone pls. tell me which one is more macthing?

Thanks

 

[extraord]

in deep submicro <0.18um first is much better, absolutly

 

[safwatonline]

Hello extraord,
can u elaborate more.
isnt both common centroid ?
as far as i know the matching happens from placement of the device so that when doped they have nearly the same effective doping ratio, which means that placment is enough for the matching , then if D and S are replaced then it doesnt matter which one is which as they r supposed to be identical ?
plz correct me if i am wrong
regards,
a.safwat

 

[Hughes]

The first one is more like a common centroid -- For the first one, the Drains are common centroid, the Sources are common centroid, and the Gates are common centroid.

The first one is better in matching, at least not worse than the second one.

 

[jcpu]

Hello:

1st one is closer to common practice.
Reason is that recently we tend to look at MOS matching
from ION implantation point o view.
This is slowly varying profile, hence fig.1 tends to average better.

However, my suggestion is
to add dummy at left and right end to avoid abnormal etching
at the boundary.

Good luck,

 

[longstar]

Jcpu is right.you can get better match if you add dummy to make the circumstance between M21 and M22 be uniform.Surely,the first is better than the second.

 

[safwatonline]

hi,
i dont see the use of a dummy we r not etching and cause the MOST most probably will have a region surrounding the active area that already acts as a dummy region

 

[pixel]

just imagine that fabrication maski is slightly shifted left(or right), and see what happens with drain and source areas for all three mosfets....
1st is better

 

[qiushidaren]

I think the first one is better.

 

[extraord]

the source and drain are not the same, in modern process.
the foundary usually do LDD in source and drain, so the plasma is not
absolut vertical. although they shift the wafer around to make the source and
drain doping similar, they are still different.

 

[vilas_nakade]

First one will give better matching because matching depends on placement and current direction. Since in first case current flow in same direction i.e. from left to right.

 

[jolia]

the first one is better

 

[leelamadhav]

The first one is better

 

[drasta]

It is true than on first currents flow in the same direction but if you shift mask I dont see how first can be better because than M21 and M22 will have smaller D than S or vice versa but on the figure 2 shifting will afect the same D and S. So be carefull.

 

[Hughes]

It is true than on first currents flow in the same direction but if you shift mask I dont see how first can be better because than M21 and M22 will have smaller D than S or vice versa but on the figure 2 shifting will afect the same D and S. So be carefull.

If you shift mask, M2 in Fig.2 will still have balanced D and S while M1 not. In Fig.1, both M1 and M2 will have unbalanced D and S. So matching is better in Fig.1.

Fig.2 has a "better matching" between S and D of M2, but not between M1 and M2.

 

[sankudey]

OK....
I have another opinion...

What we are talking abt is the averaging of variation in the horizontal direction only...
From that....I guess first one is preferable....and symmetric too....

But what abt vertical variations....u can see the area looks like rectengular....

I propose...to brak the M2 in four parts...M21...M22...M23...M24.....and put four of the transistors ..either at the diagonal of the M1 or just at horizontal/vertical axix.....generally done in BandGap reference....

May be ..it will take large are....but matcjing will be better..in my opinion....

sankudey

 

[k_90]

This is the correct way to lay out a current mirror !!!

Please find a power point presentation.